Day 2 at CEAS

By my own comparison, this year's conference was far better than last year. Last year's conference was very Bayes centric. A large portion of the papers that were presented were on different ways to use and do Bayesian analysis for spam filtering. Although interesting, one can only take so many presentations on Bayesian analysis before losing their minds. Thankfully this year's papers were far more diverse!

I probably shouldn't have written yesterday's entry as early as I did because some rather interesting things happened yesterday evening.

For starters, I had the opportunity to spend some time and talk with Wietse Venema. If you don't know who he is (I didn't when I originally met him) he is the author of the postfix MTA. Very friendly, cordial, interesting, and humble guy. It was a treat to meet him.

The shock of the day, however came when I was introduced to a gentleman named Adam Dawes from Postini. Yes, Postini! Why would I care about this? Postini has long been chastised for their lack of attendance at industry events. Whether it be MAAWG, the AOTA, or , Postini representation was nowhere to be found. It was definitely nice getting to meet Adam, and I am looking forward to seeing him or other Postini folks at future events. I am not sure if their attendance was primarily being driven by Google or if this was in the works prior to the acquisition, but either way, it is good to see them coming out.

Some more interesting sessions today, and some that although might have sounded good in a theoretical sense, would never work in a practical sense. One such example was a paper on Human Interactive Proofs for Spam Filtering. At its root, this is mostly the same as Challenge/Response (with a couple of twists). Read here for some of my feelings on C/R.

A common theme throughout several papers in the conference was around the theme of social networks. With the prevalence of sites like Facebook, MySpace, and Friendster, it only makes sense to make correlations between social networks and email patterns. You generally tend to email and communicate with certain people within closed groups. This could be the people that you work with, go to church with, or play softball with. Either way, these are closed communities within which you share some kind of common ground. The thought process being that if these people communicate with you a certain number of times and you reciprocate communication with that person, then you can make assumptions as to the legitimacy of the email that you receive from them (and vice versa).

Another paper presented by Chris Fleizach of UCSD discussed a method called Occam's Razor which is another method in which the mail recipient requests a real-time affirmation for each e-mail from the declared sender's MX. An interesting theory, but is one that without almost unanimous adoption across the industry (we know how hard THAT is to do!) wouldn't have an impact. Plus, without some big time corporate sponsor to take the banner and run with it, it's not likely to get much play within the industry.

Similar to yesterday though, the content of the conference was very good. Despite the holes that could be shot in some of the theories that were presented, it is clear that there are many innovative ideas from both industry and academia that if buttoned up could present some interesting alternative spam filtering methods. It's definitely refreshing to see this "out of the box" type of thinking because sometimes when you are so involved "in the box" it is hard to truly step outside of it and look at different ways to skin the proverbial cat.